Researchers at Osaka Metropolitan University have developed a method to detect and correct common labeling errors in large collections of radiographic images. This system automatically verifies tags related to body parts, projection, and rotation, significantly enhancing the accuracy of deep-learning models used in clinical tasks and research. The study, published in *European Radiology*, addresses challenges posed by manual labeling, which is prone to errors, especially in busy hospital settings. The team created two models—Xp-Bodypart-Checker and CXp-Projection-Rotation-Checker—both achieving high accuracy rates. Future improvements aim to further refine the model for clinical applications.
Tue, 06 Jan 2026 11:23:28 GMT | healthcare-in-europe.com